Hybrid Deep Morpho-Temporal Framework for Oscillometric Blood Pressure Measurement

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Cardiovascular Monitoring

Background:

• Oscillometric blood pressure (BP) devices are standard for automated BP measurement but have variable accuracy.
• Current algorithms fail to fully utilize the information within oscillometric pulse patterns.
• Improving BP measurement accuracy is crucial for various patient groups and settings.

Purpose of the Study:

• To develop a new 2D oscillometric data representation for enhanced BP characterization.
• To apply deep learning for extracting informative features correlated with BP from this representation.
• To improve the accuracy and reliability of automated BP measurement devices.

Main Methods:

• Proposed a novel 2D oscillometric data representation for arterial system characterization.
• Developed a hybrid convolutional-recurrent neural network (CNN-RNN) for feature extraction.
• Trained and validated the model on three oscillometric databases from 245 individuals.

Main Results:

• Achieved mean errors of 0.08 mmHg (systolic) and 0.04 mmHg (diastolic) with standard deviations of 2.4 mmHg and 2.2 mmHg, respectively.
• Outperformed existing state-of-the-art BP estimation techniques.
• Exceeded current international standards for BP monitors by a significant margin.

Conclusions:

• The proposed 2D data representation and deep learning model offer a significant advancement in BP estimation.
• This method enables more robust and objective BP monitoring across diverse patient populations and situations.
• The findings suggest a promising future for AI-driven improvements in non-specialist BP measurement tools.